Research Of Automated Visual Detection System For Foreign Object Debris On Airport Runway

Foreign Object Debris(FOD)is the debris which may damage the aircraft on the runway.FOD does not only threaten the lives of passager,but also it can result in economic losses.At present,the detection task of a vast majority of airport in global is still taken on by workers.Fortunately,as a primary method for ensuring airplane’s safety,automatical detection of Foreign Object Debris(FOD)has been a significant issue due to its advantage that the system can work all the time and under all-weather condition.In this paper,we conduct the research of automated video detection and surveillance system for foreign object debris on airport runway,with help of the official requirements of FOD detection performance and all related studies before in the field.And three difficulties are identified: the hardware design of optical detector in ultra distance,FOD image extraction algorithm and FOD recognition algorithm.We focus on these difficulties and take a amount of effort to solve them.Finally,all efforts are paid off,and the prototype has been deployed and verified at the Shanghai Hongqiao International Airport.In this paper,the main contribution in hardware design and software algorithm can be summarized as follows:1.The paper accomplishes the design of image system and mechanical system of optical detector.We conduct a series of experiments about the install location of detector,and based on all the work,the final designing scheme is worked out and verified.In addition,the paper gives the standard work processes of parameter calculation and product selection of hardware.2.The paper proposes a customized approach of object recognition in FOD visual system aiming at airfield runway.It is designed to correct both optical and perspective distortions,and get rid of the variation of ambient light.The approach has been tested in the surveillance image database,and the experimental results show that it performs high accuracy,computing efficiency and robustness.Its correct recognition rate reaches up to 95%.